[FIXED] Matplotlib: Plot circles and scale them up so the text inside doesn't go out of the circle bounds

Issue

I have some data where i have languages and relative unit size. I want to produce a bubble plot and then export it to PGF. I got most of my code from this answer Making a non-overlapping bubble chart in Matplotlib (circle packing) but I am having the problem that my text exits the circle boundary:

How can I either, increase the scale of everything (much easier I assume), or make sure that the bubble size is always greater than the text inside (and the bubbles are still proportional to each other according to the data series). I assume this is much more difficult to do but I don't really need that.

Relevant code:

#!/usr/bin/env python3
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.colors as mcolors

# create 10 circles with different radii
r = np.random.randint(5,15, size=10)
mapping = [("English", 25),
                    ("French", 13),
                    ("Spanish", 32),
                    ("Thai", 10),
                    ("Vientamese", 13),
                    ("Chinese", 20),
                    ("Jamaican", 8),
                    ("Scottish", 3),
                    ("Irish", 12),
                    ("American", 5),
                    ("Romanian", 3),
                    ("Dutch", 2)]

class C():
    def __init__(self,r):
        self.colors = list(mcolors.XKCD_COLORS)
        self.N = len(r)
        self.labels = [item[0] for item in r]
        self.x = np.ones((self.N,3))
        self.x[:,2] = [item[1] for item in r]
        maxstep = 2*self.x[:,2].max()
        length = np.ceil(np.sqrt(self.N))
        grid = np.arange(0,length*maxstep,maxstep)
        gx,gy = np.meshgrid(grid,grid)
        self.x[:,0] = gx.flatten()[:self.N]
        self.x[:,1] = gy.flatten()[:self.N]
        self.x[:,:2] = self.x[:,:2] - np.mean(self.x[:,:2], axis=0)

        self.step = self.x[:,2].min()
        self.p = lambda x,y: np.sum((x**2+y**2)**2)
        self.E = self.energy()
        self.iter = 1.

    def minimize(self):
        while self.iter < 1000*self.N:
            for i in range(self.N):
                rand = np.random.randn(2)*self.step/self.iter
                self.x[i,:2] += rand
                e = self.energy()
                if (e < self.E and self.isvalid(i)):
                    self.E = e
                    self.iter = 1.
                else:
                    self.x[i,:2] -= rand
                    self.iter += 1.

    def energy(self):
        return self.p(self.x[:,0], self.x[:,1])

    def distance(self,x1,x2):
        return np.sqrt((x1[0]-x2[0])**2+(x1[1]-x2[1])**2)-x1[2]-x2[2]

    def isvalid(self, i):
        for j in range(self.N):
            if i!=j:
                if self.distance(self.x[i,:], self.x[j,:]) < 0:
                    return False
        return True

    def scale(self, size):
        """Scales up the plot"""
        self.x = self.x*size

    def plot(self, ax):
        for i in range(self.N):
            circ = plt.Circle(self.x[i,:2],self.x[i,2], color=mcolors.XKCD_COLORS[self.colors[i]])
            ax.add_patch(circ)
            ax.text(self.x[i][0],self.x[i][1], self.labels[i], horizontalalignment='center', size='medium', color='black', weight='semibold')

c = C(mapping)

fig, ax = plt.subplots(subplot_kw=dict(aspect="equal"))
ax.axis("off")

c.minimize()

c.plot(ax)
ax.relim()
ax.autoscale_view()
plt.show()

Solution

I think both approaches that you outline are largely equivalent. In both cases, you have to determine the sizes of your text boxes in relation to the sizes of the circles. Getting precise bounding boxes for matplotlib text objects is tricky business, as rendering text objects is done by the backend, not matplotlib itself. So you have to render the text object, get its bounding box, compute the ratio between current and desired bounds, remove the text object, and finally re-render the text rescaled by the previously computed ratio. And since the bounding box computation and hence the rescaling is wildly inaccurate for very small and very large text objects, you actually have to repeat the process several times (below I am doing it twice, which is the minimum).

W.r.t. the placement of the circles, I have also taken the liberty of substituting your random walk in an energy landscape with a proper minimization. It's faster, and I think the results are much better.

#!/usr/bin/env python3
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.colors as mcolors

from scipy.optimize import minimize, NonlinearConstraint
from scipy.spatial.distance import pdist, squareform


def _get_fontsize(size, label, ax, *args, **kwargs):
    """Given a circle, precompute the fontsize for a text object such that it fits the circle snuggly.

    Parameters
    ----------
    size : float
        The radius of the circle.
    label : str
        The string.
    ax : matplotlib.axis object
        The matplotlib axis.
    *args, **kwargs
        Passed to ax.text().

    Returns
    -------
    fontsize : float
        The estimated fontsize.
    """

    default_fontsize = kwargs.setdefault('size', plt.rcParams['font.size'])
    width, height = _get_text_object_dimensions(ax, label, *args, **kwargs)
    initial_estimate = size / (np.sqrt(width**2 + height**2) / 2) * default_fontsize
    kwargs['size'] = initial_estimate
    # Repeat process as bbox estimates are bad for very small and very large bboxes.
    width, height = _get_text_object_dimensions(ax, label, *args, **kwargs)
    return size / (np.sqrt(width**2 + height**2) / 2) * initial_estimate


def _get_text_object_dimensions(ax, string, *args, **kwargs):
    """Precompute the dimensions of a text object on a given axis in data coordinates.

    Parameters
    ----------
    ax : matplotlib.axis object
        The matplotlib axis.
    string : str
        The string.
    *args, **kwargs
        Passed to ax.text().

    Returns
    -------
    width, height : float
        The dimensions of the text box in data units.
    """

    text_object = ax.text(0., 0., string, *args, **kwargs)
    renderer = _find_renderer(text_object.get_figure())
    bbox_in_display_coordinates = text_object.get_window_extent(renderer)
    bbox_in_data_coordinates = bbox_in_display_coordinates.transformed(ax.transData.inverted())
    w, h = bbox_in_data_coordinates.width, bbox_in_data_coordinates.height
    text_object.remove()
    return w, h


def _find_renderer(fig):
    """
    Return the renderer for a given matplotlib figure.

    Notes
    -----
    Adapted from https://stackoverflow.com/a/22689498/2912349
    """

    if hasattr(fig.canvas, "get_renderer"):
        # Some backends, such as TkAgg, have the get_renderer method, which
        # makes this easy.
        renderer = fig.canvas.get_renderer()
    else:
        # Other backends do not have the get_renderer method, so we have a work
        # around to find the renderer. Print the figure to a temporary file
        # object, and then grab the renderer that was used.
        # (I stole this trick from the matplotlib backend_bases.py
        # print_figure() method.)
        import io
        fig.canvas.print_pdf(io.BytesIO())
        renderer = fig._cachedRenderer
    return(renderer)


class BubbleChart:

    def __init__(self, sizes, colors, labels, ax=None, **font_kwargs):
        # TODO: input sanitation

        self.sizes = np.array(sizes)
        self.labels = labels
        self.colors = colors
        self.ax = ax if ax else plt.gca()

        self.positions = self._initialize_positions(self.sizes)
        self.positions = self._optimize_positions(self.positions, self.sizes)
        self._plot_bubbles(self.positions, self.sizes, self.colors, self.ax)

        # NB: axis limits have to be finalized before computing fontsizes
        self._rescale_axis(self.ax)

        self._plot_labels(self.positions, self.sizes, self.labels, self.ax, **font_kwargs)


    def _initialize_positions(self, sizes):
        # TODO: try different strategies; set initial positions to lie
        # - on a circle
        # - on concentric shells, larger bubbles on the outside
        return np.random.rand(len(sizes), 2) * np.min(sizes)


    def _optimize_positions(self, positions, sizes):
        # Adapted from: https://stackoverflow.com/a/73353731/2912349

        def cost_function(new_positions, old_positions):
            return np.sum((new_positions.reshape((-1, 2)) - old_positions)**2)

        def constraint_function(x):
            x = np.reshape(x, (-1, 2))
            return pdist(x)

        lower_bounds = sizes[np.newaxis, :] + sizes[:, np.newaxis]
        lower_bounds -= np.diag(np.diag(lower_bounds)) # squareform requires zeros on diagonal
        lower_bounds = squareform(lower_bounds)

        nonlinear_constraint = NonlinearConstraint(constraint_function, lower_bounds, np.inf, jac='2-point')
        result = minimize(lambda x: cost_function(x, positions), positions.flatten(), method='SLSQP',
                        jac="2-point", constraints=[nonlinear_constraint])
        return result.x.reshape((-1, 2))


    def _plot_bubbles(self, positions, sizes, colors, ax):
        for (x, y), radius, color in zip(positions, sizes, colors):
            ax.add_patch(plt.Circle((x, y), radius, color=color))


    def _rescale_axis(self, ax):
        ax.relim()
        ax.autoscale_view()
        ax.get_figure().canvas.draw()


    def _plot_labels(self, positions, sizes, labels, ax, **font_kwargs):
        font_kwargs.setdefault('horizontalalignment', 'center')
        font_kwargs.setdefault('verticalalignment', 'center')

        for (x, y), label, size in zip(positions, labels, sizes):
            fontsize = _get_fontsize(size, label, ax, **font_kwargs)
            ax.text(x, y, label, size=fontsize, **font_kwargs)


if __name__ == '__main__':

    mapping = [("English", 25),
               ("French", 13),
               ("Spanish", 32),
               ("Thai", 10),
               ("Vietnamese", 13),
               ("Chinese", 20),
               ("Jamaican", 8),
               ("Scottish", 3),
               ("Irish", 12),
               ("American", 5),
               ("Romanian", 3),
               ("Dutch", 2)]

    labels = [item[0] for item in mapping]
    sizes = [item[1] for item in mapping]
    colors = list(mcolors.XKCD_COLORS)

    fig, ax = plt.subplots(figsize=(10, 10), subplot_kw=dict(aspect="equal"))
    bc = BubbleChart(sizes, colors, labels, ax=ax)
    ax.axis("off")
    plt.show()

Answered By - Paul Brodersen